Incontro IDEE Meeting

1.

Speakers: Mathilde Lesueur and Pablo Alvarez Aragon

Title: Measuring Attitudes Toward Harsh Parenting: Designing a Vignette-Based Factorial Survey for Italy

Abstract: As part of the ERC-funded HARSH project, we are developing a novel survey instrument to measure attitudes and perceived social norms regarding violence in child discipline. The core of the instrument is a factorial survey experiment (vignette) in which respondents evaluate parental reactions to child misbehaviour. In this talk, we will present the draft questionnaire and discuss open design choices where we would especially welcome feedback: the wording and realism of the vignette scenarios, the trade-off between respondent burden and experimental power, the measurement of second-order beliefs, sampling design and representativeness across demographic and geographic characteristics, and strategies to mitigate social desirability bias — including the use of a randomized face-saving prologue and the placement of sensitive retrospective questions on childhood experiences.

2.

Speaker: Roujing Wu

Title: Salience Distraction: Evidence From LLM-Measured Exam Shocks

Abstract: This paper studies how exam design–induced shocks affect performance in a high‑stakes real-world setting. Using administrative data on all Gaokao takers in China from 1999–2003, matched to digitized math papers, we employ an LLM to construct question‑level difficulty measures and treat an unexpectedly hard first question as a negative shock. Conditional on overall exam difficulty, a harder first question significantly lowers math scores, with disproportionately larger losses for higher‑score students. This pattern is consistent with a salience‑distraction mechanism: an abnormally difficult item in a prominent position captures attention and crowds out cognitive resources needed for the rest of the test. Heterogeneity analyses further show that students from high‑quality high schools and those in provincial capital cities are substantially less exposed to this performance loss, suggesting that richer resources and stronger preparation buffer the impact of negative shocks. Overall, while aggregate sorting by ability appears stable, exam‑design shocks generate unequal behavioral penalties at the individual level.